From c38a79880d814041b93e1e94ca39d4e0fa0727a1 Mon Sep 17 00:00:00 2001 From: Jacky <18255193+kthui@users.noreply.github.com> Date: Tue, 3 Dec 2024 16:20:45 -0800 Subject: [PATCH] [docs] Minor wording updates Co-authored-by: Ryan McCormick --- docs/additional_outputs.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/additional_outputs.md b/docs/additional_outputs.md index 4c531a1..1cb9e13 100644 --- a/docs/additional_outputs.md +++ b/docs/additional_outputs.md @@ -57,7 +57,7 @@ point value will be sent on the `cumulative_logprob` output tensor. ### Log Probabilities -The log probabilities of the top probability words at each position if the +The log probabilities of the top probability tokens at each position of the [logprobs](https://github.com/vllm-project/vllm/blob/v0.6.3.post1/vllm/sampling_params.py#L146-L152) are requested. Only the log probabilities of the new tokens generated since the last response are returned on each new response. See @@ -65,7 +65,7 @@ last response are returned on each new response. See for more details on the log probabilities. To enable, set `return_logprobs` input tensor to `True`. The log probabilities -will be sent on the `logprobs` output tensor as JSON. +will be sent on the `logprobs` output tensor as a serialized JSON string. ### Number of Input Tokens